Triple-helical nucleic acids are formed by binding an oligonucleotide within the major groove of duplex DNA. These complexes offer the possibility of designing oligonucleotides which bind to duplex DNA with considerable sequence specificity. However, triple-helix formation with natural nucleotides is limited by (i) the requirement for low pH, (ii) the requirement for homopurine target sequences, and (iii) their relatively low affinity. We have prepared modified oligonucleotides to overcome these limitations, including the addition of positive charges to the sugar and/or base, the inclusion of cytosine analogues, the development of nucleosides for recognition of pyrimidine interruptions and the attachment of one or more cross-linking groups. By these means we are able to generate triplexes which have high affinities at physiological pH at sequences that contain pyrimidine interruptions.
- bis-amino-U (BAU)
- DNA recognition
- nucleotide analogue
- triple-helix formation
Triple-helical nucleic acids can be generated by binding an oligonucleotide within the major groove of duplex DNA and are stabilized by the formation of hydrogen bonds between the third-strand bases and exposed groups on the purine strand of duplex DNA. The third strand can be oriented in either the parallel or anti-parallel direction, relative to the purine strand, although parallel triplexes are usually more stable, especially at low pH. In the parallel motif, GC is recognized by protonated cytosine, while AT is bound by thymine, generating C+.GC and T.AT triplets (Figure 1A). These structures, which were first noted with polynucleotides in 1957 , and subsequently with TFOs (triplex-forming oligonucleotides) targeted to sites within duplex DNA in 1987 [2,3], offer the possibility of designing oligonucleotides which bind to duplex DNA with considerable sequence specificity. However, triple-helix formation with natural nucleotides is limited by (i) the requirement for low pH (necessary for formation of the C+.GC triplet), (ii) the requirement for homopurine target sequences, and (iii) the relatively low affinity due to charge repulsion between the polyanionic strands. We have prepared heavily modified oligonucleotides (while retaining the phosphodiester backbone) in order to overcome these limitations, which include the addition of positive charges to the sugar and/or base, the inclusion of cytosine analogues, the development of nucleosides for recognition of pyrimidine interruptions and the attachment of one or more cross-linking groups (psoralen). By these means we are able to generate triplexes, which have high affinities at physiological pH at sequences that contain pyrimidine interruptions.
Nucleic acids triple-helices are usually much less stable than their duplex counterparts on account of their slow rate of formation (typically three orders of magnitude slower than duplexes [4,5]) and electrostatic repulsion between the three polyanionic strands. The T.AT triplet is less stable than C+.GC [6–9] and requires millimolar concentrations of magnesium for effective formation . Several strategies have been employed to enhance the stability of this triplet, including increasing the aromatic surface area  and altering the backbone conformation to N3′–P5′ , PNA (peptide nucleic acid) [12,13] or LNA (locked nucleic acid) [14,15] (third strands with N-type sugar conformations are more stable than S-type and RNA-like third strands bind well to duplex DNA targets [16,17]). We have sought to increase stability by the addition of positively charged groups to the sugar, the base or both. 5-Propargylamino-dU (Figure 1B)  produces a large increase in affinity (ΔTm≈4°C per substitution), similar to that of 2′-aminoethoxy-T , while combining both modifications in the analogue BAU (bis-amino-U) produces a dramatic increase in stability with an increase in Tm of up to 8°C per substitution [20–22]. The additive effect of these two substitutions occurs because they contact different phosphates; the 2′-aminoethoxy group interacts with the adjacent phosphate in the duplex purine strand, whereas the 5-substitution binds to a phosphate in the TFO itself. Triplexes formed with these charged analogues no longer require the addition of divalent cations.
Inclusion of BAU has a large effect on triplex kinetics, and melting curves with TFOs that contain only a single substitution show considerable hysteresis, even at slow rates of heating and cooling [20,21]. Analysis of the hysteresis profiles reveals that even a single substitution with this analogue within an 18-mer TFO produces a 10-fold decrease in the rate of dissociation, with little change in the association rate. Inclusion of a second BAU residue further enhances the stability and reduces the rate of dissociation. In this instance, the effect is greatest when the two modified nucleotides are closer together . The selectivity of BAU for AT base pairs is also reflected in a much slower rate of dissociation from this compared with other base pairs.
We have used BAU in combination with other derivatives, described below, to achieve triplex recognition of mixed DNA sequences at physiological pH [23,24] and it is the most stabilizing derivative that we have developed. However, synthesis of this phosphoramidite is not trivial and, as far as we are aware, this analogue has not subsequently been used by other groups. In order to simplify the synthesis we have examined the effects of dimethylaminopropargyl-dU on triplex stability  (since this does not need protecting groups during oligonucleotide synthesis and it can be used with other derivatives that require mild deprotection conditions). This derivative still enhances triplex stability, although to a lesser extent than other analogues that bear positive charges. This represents a useful nucleotide analogue that combines the ease of synthesis with triplex stabilization.
For oligonucleotides that contain multiple substitutions with these analogues, their positions and arrangement within the TFO are important. For multiple additions of BAU, clustering these towards one end of the TFO produced a much smaller increase in affinity than when they were evenly distributed throughout the oligonucleotide . A similar effect has been noted with the dimethylaminopropargyl derivative , although in this case the lower stabilization of adjacent resides may be the result of steric clash between methyl groups on adjacent residues.
In some instances, these modifications produce triplexes that are now more stable than the underlying duplex and the apparent Tm in melting experiments is limited by the duplex stability . In these instances, when working with short duplex targets, the apparent triplex stability depends on the sequence and length of the flanking duplex sequences. For a 14-mer parallel TFO, the Tm at pH 5.0 was 10°C lower with an intermolecular 14-mer duplex target than with an intramolecular duplex, or one that was flanked by six GC base pairs at either end. The use of simple intermolecular duplex targets may therefore underestimate the triplex stabilization that is produced by some nucleotide analogues. Target sites that are contained within longer, more stable, duplexes will also more closely reflect the situation in vivo when the targets will be flanked by the remainder of the genomic DNA.
The addition of intercalating or cross-linking groups to TFOs has also been used to increase TFO affinity. These groups are usually attached to the 5′-end of the TFO, although they can also be added to the 3′- or both ends, generating a triplex staple. Intercalating compounds such as acridine or anthraquinone can be attached at either or both ends of the oligonucleotide. Addition of intercalators at both 3′- and 5′-ends of an oligonucleotide can produce triplexes with melting temperatures above 40°C at pH 7 even though they do not contain any triplex-stabilizing base analogues . It has been suggested that 5′-attachment of an intercalators enhances stability more than 3′-attachment, but this may simply reflect the observation that intercalation at YpR is favoured over RpY. Psoralen can also be attached at either end or at an internal site of a TFO and UV irradiation of these complexes produces interstrand cross-links at adjacent TpA steps. Photoreaction of TFOs containing two psoralens, located at the 5′- and 3′-ends, has been used to create cross-links at both termini of the TFO (triplex staples), producing complexes that have no free single-stranded ends .
Secondary binding sites
We have examined whether the addition of positive charges affects the stringency of triple-helix formation. TFOs with single substitutions of BAU bind best when this base is placed opposite AT as expected . The next best combination is BAU.GC, and although this is more stable than T.GC, the discrimination between AT and GC is not affected. Triplets with BAU placed opposite TA or CG are not stabilized relative to T.CG and T.TA, demonstrating that BAU has enhanced discrimination for AT (and GC) over CG and TA relative to T. Oligonucleotides that contain multiple substitutions with these charged analogues do show some weaker interaction with secondary sites. In particular, TFOs that contain (BAU)n tracts can cause the oligonucleotide to bind to An.Tn tracts even though the surrounding duplex sequence does not match the remainder of the TFO sequence (A. Cardew, T. Brown and K.R. Fox, unpublished work).
The pH-dependence of the C+.GC triplet presents a major limitation to triplex formation at physiological conditions, which is especially pronounced when recognizing adjacent GC base pairs . A large number of cytosine analogues with elevated pKa values have been developed to overcome this problem. Early studies used 5-methylcytosine, which has a slightly higher pKa than C , although the improved binding probably arises from the stacking of the additional methyl groups within the major groove (in the same way as T.AT is more stable than U.AT). A selection of analogues that have been examined is presented in Figure 2. The ones that are based on the pyrimidine nucleus are generally isostructural with C and form triplets that are isomorphic with T.AT. Of these analogues, only 2-aminopyridine [31–33] retains the positive charge (the nucleoside has a pKa of 5.93 compared with 4.3 for dC) and, unlike the others, it produces triplexes that are as stable as those formed with C at low pH. Those lacking the stabilizing positive charge may need to be used in combination with other stabilizing analogues, such as 2′-aminoethoxy derivatives . For the analogues that are based on a purine nucleus the glycosidic bond is in a different position, causing distortions in the backbone between X.GC and T.AT triplets. These are therefore best used for recognizing longer G-tracts in a pH-independent fashion. The affinity of C for GC has also been enhanced by appending positively charged groups such as spermine  to the base or by attaching it to 2′-OMe (2′-O-methyl) groups or 2′-aminoethoxyribose sugars [19,36]. These modifications compensate for the lower affinity of unprotonated cytosine. Further increases in stability, extending triplex formation to neutral pH, can be produced by combining these analogues with other base analogues such as 5-propynyl-dU or BAU [23,34,37]. We have successfully used 2-aminopyridine (and its 3-methyl derivative) together with modification at the 2′-position of the sugar to produce triplexes that are stable at higher pHs (C. Lou, K.R. Fox and T. Brown, unpublished work). When combined with BAU, this can be used to produce triplexes that form at physiological pH with submicromolar dissociation constants.
Since the third strand only makes contacts with the purine base of each base pair, triple-helix formation is generally restricted to homopurine sequences; even single pyrimidine interruptions can abolish triplex formation. Targeting pyrimidine bases is not simple as they present fewer hydrogen-bonding points than purines, and there is some steric interference from the 5-methyl group of thymine. A number of synthetic nucleotides have been developed to overcome this restriction and many of these make hydrogen-bond contacts with both bases of the YR base pair. The synthetic nucleobase S (Figure 3)  and its 2′-aminoethoxy derivative  is one of the most effective for recognition of TpA, although this has low-sequence discrimination and also binds well to CG, GC and AT. Although it has lower selectivity than formation of the G.TA triplet (the best combination of natural nucleotides for recognition of TA) it binds across this pyrimidine interruption better than any other combination. Indeed, this nucleobase can also be used for producing stable complexes across CG interruptions. We have developed a series of synthetic nucleobases, based on substituted 3H-pyrrolo[2,3-d]pyrimidin-2(7H)-one nucleosides, for recognizing CG inversions (see Figure 3) [40–42]. These generally give good selectivity and affinity, but the relative stability can be affected by the surrounding sequence.
One goal of triplex technology is to be able to target any sequence with high affinity at physiological pH. The development of derivatives of C and T for improved recognition of AT and GC means that we are now able to generate high-affinity triplexes at most oligopurine sites at physiological pH. However, the presence of pyrimidine interruptions in the target still presents a challenge. Targets that contain single YR interruptions can be efficiently targeted by using the S or APP derivatives in combination with the high-affinity T and C derivatives. This strategy can produce triplexes with low nanomolar dissociation constants at physiological pH. TFOs for targets that contain several pyrimidine interruptions have lower affinity. However, by the combined use of BAU (for strong binding to AT), aminopyridine (for the recognition of GC at high pH), S (for recognizing TA, albeit with low selectivity) and APP (for recognition of CG) we have successfully targeted a 19-mer oligopurine sequence containing two CG and two TA interruptions, with micromolar affinity at physiological pH .
This work was supported by grants from the Biotechnology and Biological Sciences Research Council.
We thank Chenguang Lou (Harvey), Radha Taylor, Antonia Cardew, David Rusling, Simon Gerrard, Imenne Bouamaied and Nuria Vergara for their contributions to this work.
NACON VIII: 8th International Meeting on Recognition Studies in Nucleic Acids: An Independent Meeting held at The Edge, University of Sheffield, Sheffield, U.K., 12–16 September 2010. Organized by Mike Blackburn, Mark Dickman, Jane Grasby, David Hornby, Chris Hunter, John Rafferty, Jim Thomas, David Williams and Nick Williams (Sheffield, U.K.).
Abbreviations: BAU, bis-amino-U; TFO, triplex-forming oligonucleotide
- © The Authors Journal compilation © 2011 Biochemical Society